Overview of Capital Account Crisis - IMF

INTERNATIONAL MONETARY FUND
Capital Account Crises: Lessons for Crisis Prevention
Atish Ghosh 1
July 2006
I. INTRODUCTION
1. The financial crises that struck a number of emerging market countries in the 1990s
and early twenty-first century were characterized by sudden reversals of capital flows that
had pervasive macroeconomic consequences, including abrupt current account adjustment
and collapsing real exchange rates and economic activity (Figure 1). But while the
consequences of these crises were broadly similar, their causes appear to be bewilderingly
different. Turkey (1993), Mexico (1994), and Russia (1998) were public sector funding
crises. By contrast, the 1997 East Asian crises were mainly private sector phenomena. In
Brazil (1998-99), Turkey (2000-01) and Argentina (2002) public sector debt dynamics
played a key role—in the latter two cases, accompanied by a banking crisis. On the other
hand, Uruguay (2002) was a banking crisis—caused by withdrawals of Argentine deposits—
that spilled into a public sector debt problem and a balance of payments crisis.
2. Nor has the academic literature been able to give a coherent and unified account that
explains all of these crises. The first generation of currency crisis models (Krugman 1979,
Flood and Garber 1984) emphasized the inconsistency between financing a budget deficit
through money creation and trying to maintain a pegged exchange rate regime. Since these
models did not seem to fit the 1992/93 European Exchange Rate Mechanism crises, a second
generation of crisis models (Obstfeld 1994) was developed in which an inconsistent policy
stance, combined with self-fulfilling shifts in investor sentiments, could give rise to multiple
equilibria. Yet neither variant could explain the East Asian crises, necessitating a third
generation that incorporated foreign exchange exposure of the private financial and corporate
sectors. But this third generation of currency crisis models could not explain subsequent
crises, such as Argentina (2002). And while the collapse of Argentina’s currency board
1 Prepared for the High Level Seminar on Crisis Prevention, Singapore, July 10-11, 2006.
The author is Division Chief, Policy Development and Review Department, IMF. The views
expressed in this paper are those of the author and should not be attributed to the
International Monetary Fund, its Executive Board, or its Management.

- 2 -
Figure 1. Selected Macroeconomic Indicators: Mean 1/
12
Real GDP growth
(In percent per year)
40
Inflation
(In percent per year)
9
35
30
6
25
3
20
0
15
-3
10
5
-6
0
-9
t-3 t-2 t-1 t t+1 t+2 t+3
-5
t-3 t-2 t-1 t t+1 t+2 t+3
20
Real exchange rate
(Annual change, in percent)
15
Current account balance
(In percent of GDP)
10
10
0
5
-10
0
-20
-5
-30
t-3 t-2 t-1 t t+1 t+2 t+3
-10
t-3 t-2 t-1 t t+1 t+2 t+3
Sources: International Monetary Fund; WEO database, and IMF staff estimates.
1/ Averages (Mean) are given by the solid lines, with standard deviations around the mean given by the
dotted lines. The sample consists of Argentina (1995 and 2002), Brazil (1999), Indonesia (1997), Malaysia
(1997), Mexico (1997), Philippines (1997), Russia (1998), Thailand (1997), Turkey (2000), and Uruguay
(2002).

- 3 -
resulted mainly from a fiscal policy stance that was incompatible with the exchange rate
regime, the crisis was not in the mold of the first generation models as the government was
bond financing its deficit in a deflationary, rather than an inflationary, environment. 2
3. All this suggests that understanding capital account crises—surely a prerequisite to
preventing them—requires a more general analytical framework. The central thesis of this
paper is that a capital account crisis requires—and is caused by—a combination of balance
sheet weaknesses in the economy and a specific crisis trigger. The diversity of capital
account crises is therefore not surprising because balance sheet weaknesses can take various
forms, as can the specific factors that trigger the crisis. Much like a bomb that requires both
an explosive material and a detonator to cause an explosion, neither the balance sheet
weakness nor the crisis trigger on its own is likely to cause (as much) mischief. Thus an
economy can live with currency and maturity mismatches in private or public sectoral
balance sheets for years if, serendipitously, nothing triggers a crisis. Yet there are many
possible crisis triggers, both external—contagion, a terms of trade shock, a deterioration in
market conditions—and domestic, such as an inconsistent macroeconomic policy stance (see
Table 1 for a summary of vulnerabilities and crisis triggers in selected emerging market
countries).
4. Since emerging market countries still typically lack the ability to borrow in their own
currencies (especially at long maturities), some currency and maturity mismatches may be
unavoidable. 3 In the same vein, while sound macroeconomic policies can help avoid certain
crisis triggers, others may be beyond the control of the country. Therefore, national
authorities should seek to avoid both balance sheet weaknesses and poor policies in order to
minimize the likelihood of a crisis.
5. The remainder of this paper is organized as follows. Drawing on recent work
undertaken at the IMF, section II provides a few illustrative examples of how interactions
between crisis triggers and underlying balance sheet vulnerabilities resulted in some of the
recent capital account crises. 4 Section III draws some general lessons for reducing balance
sheet vulnerabilities. Section IV turns to crisis prevention more generally, including some of
the possible roles of the IMF, highlighting measures that have been taken at the Fund since
the mid-1990s in this direction. Section V concludes.
2 See Box 2.1 of Roubini and Setser (2005) for a comparison of assumptions in different
generations of models.
3 Some countries, however, may find it rational to borrow in foreign currencies, given trend
real appreciation of their currencies leading to low (or even negative) real interest rates. See
Lipschitz et al. (2005) for a discussion of this case.
4 This section draws heavily on Allen et al. (2000), and Rosenberg et al. (2005).

- 4 -
Table 1. Taxonomy of Vulnerability and Triggers in Recent Capital Account Crises
Crisis Balance sheet vulnerability Crisis trigger
Mexico (1994)
Argentina (1995)
Thailand (1997)
Korea (1997)
Indonesia (1997)
Government's short-term external
(and FX-denominated) liabilities
Banking system short-term
external and peso and FXdenominated
liabilities
Financial and non-financial
corporate sector external
liabilities; concentrated exposure
of finance companies to property
sector
Financial sector external
liabilities (with substantial
maturity mismatch) and
concentrated exposure to
chaebols; high corporate
debt/equity ratio
Corporate sector external
liabilities; concentration of
banking system assets in real
estate/property-related lending;
high corporate debt/equity ratio
Tightening U.S. monetary policy;
political shocks (Chiapas;
assassination of the presidential
candidate)
Mexican ("Tequila") crisis
Terms of trade deterioration; asset
price deflation.
Terms of trade deterioration; falling
profitability of chaebols; contagion
from Thailand's crisis
Contagion from Thailand's crisis;
banking crisis
Russia (1998)
Government's short-term external
financing needs
Failure to implement budget deficit
targets; terms of trade deterioration
Brazil (1999)
Government's short-term external
liabilities
Doubts about ability to implement
budget cuts and loose budget proposal
for 1999; current account deficit;
contagion from Russian default
Turkey (2000)
Government short-term
liabilities, banking system FXand
maturity mismatches
Widening current account deficit, real
exchange rate appreciation, terms of
trade shock; uncertainty about
political will of government to
undertake reforms in the financial
sector.
Argentina (2002)
Public and private sector external
and FX-denominated liabilities.
Persistent failure to implement budget
deficit targets; inconsistency between
currency board arrangement and
fiscal policy; Russian default
Uruguay (2002)
Banking system short-term
external liabilities.
Argentine deposit freeze leading to
mass withdrawls from Uruguay

- 5 -
II. BALANCE SHEET VULNERABILITIES AND CRISIS TRIGGERS—SOME ILLUSTRATIVE
EXAMPLES
6. Traditional flow-based analysis focuses on the gradual build up of unsustainable
budget and current account deficits. The balance sheet approach (BSA) complements such
analysis by considering how shocks to stocks of assets and liabilities in sectoral balance
sheets can lead to large adjustments that are manifested in capital outflows (and
corresponding current account surpluses as external financing is withdrawn).
7. While further disaggregation is possible, BSA typically analyzes four main sectoral
balance sheets: the government sector (including the central bank), the private financial
sector, the private non-financial sector (households and corporations), and the external sector
(or “rest of the world”). This sectoral decomposition can reveal important vulnerabilities that
are hidden when considering the country’s consolidated balance sheet (or its net position visà-vis
the rest of the world). In particular, weaknesses in one sectoral balance sheet may
interact with others, eventually spilling into a country-wide balance of payments crisis even
though the original mismatch was not evident in the country’s aggregate balance sheet. A
prime example is the foreign currency debt between residents, which of course gets netted
out of the aggregate balance sheet, but may nevertheless contribute to a balance of payments
crisis. For example, if the government has foreign currency debt to residents and faces a
funding crisis, it will need to draw down the central bank’s foreign exchange reserves,
possibly leading to a balance of payments crisis.
8. More generally, a loss of confidence or a re-evaluation of risks in one sector can
prompt sudden and large scale portfolio adjustments, such as massive withdrawals of bank
deposits, panic sales of securities, or abrupt halts in debt rollovers. As the exchange rate,
interest rates, and other prices adjust, other balance sheets can sharply deteriorate, in turn
provoking creditors to shift toward safer foreign assets—resulting in capital outflows and
further pressure on the exchange rate and reserves until there is a full-blown capital account
crisis.
9. The following examples show how weaknesses in sectoral balance sheets—currency
and maturity mismatches, capital structure, and solvency—together with specific “triggers”
resulted in some of the recent capital account crises. 5 While there is undoubtedly an element
of “ex post rationalization” in identifying the crisis triggers, these examples are nevertheless
useful in illustrating how exposures in different sectoral balance sheets can interact to
produce vulnerabilities.
5 The three examples—Thailand (1997), Argentina (2002), and Turkey (2000/2001)—are
chosen from the list in Table 1 to represent three different sources of balance sheet
vulnerabilties.

- 6 -
Thailand (1997)
10. Thailand’s devaluation on July 2, 1997 was the first in a wave of capital account
crises that afflicted East Asia, eventually engulfing Korea, Indonesia, Malaysia, and the
Philippines. The macroeconomic consequences for Thailand were pervasive, with real GDP
growth falling from 9 percent in 1994/95 to -11 percent in 1998, the current account
swinging from a deficit of 8 percent of GDP in 1996 to a surplus of 13 percent of GDP
in 1998, and external debt rising from 60 percent of GDP at end-1996 to 94 percent of GDP
by end-1998.
Balance sheet vulnerabilities
11. What were the underlying balance sheet vulnerabilities? Although available data are
incomplete, Table 2 provides a snapshot (as of end-1996) of the main sectoral liabilities.
Table 2. Thailand: Sectoral Foreign Assets and Liabilities, end-1996
(In billions of U.S. dollars)
Assets Liabilities
Net
General government 38.7 5.2 33.5
Short-term 38.7 0.0 38.7
Medium- and long-term 0.0 5.1 -5.1
Commercial banks 7.0 48.1 -41.1
Short-term 2.6 28.2 -25.6
Medium- and long-term 4.4 19.9 -15.5
Domestic FX 0.0 0.0 0.0
Non-banks 0.5 98.0 -97.5
Short-term 0.5 23.6 -23.1
Medium- and long-term 0.0 42.9 -42.9
Domestic FX 0.0 31.5 -31.5
Subtotal
Short-term 41.8 51.8 -10.0
Medium- and long-term and domestic FX 4.4 99.5 -95.0
Total 46.2 151.3 -105.1
Source: Figures 1 and 2 of Allen et al. (2002).
• Thailand’s short-term liability position vis-à-vis the rest of the world was US$10 billion,
but this masked the huge currency and maturity mismatches of the banking and nonfinancial
sectors.
• Short-term net foreign liabilities of the banking system were US$25.6 billion (=US$28.2-
US$2.6 billion). Even if some of its medium- and long-term assets (US$4.4 billion) could

- 7 -
be made liquid, there remained a potential financing gap of US$21 billion if short-term
liabilities could not be rolled over.
• Of the non-bank sector’s total liabilities, some US$66.4 billion was owed to foreigners in
foreign currency (including equity, which would likely be converted into foreign
currency if foreigners withdrew), of which US$23.6 billion was short-term.
• Commercial banks were covering their overall (short- and long-term) FX-denominated
liabilities of US$48.1 billion 6 with foreign assets of US$7.0 billion and FX-denominated
claims on domestic residents of US$31.5 billion, leaving a net FX liability position of
US$9.6 billion. However, this assumed that domestic residents would be able to cover the
US$31.5 billion of FX-liabilities in the event of a devaluation. The non-financial sector’s
foreign liabilities amounted to US$98 billion (against foreign asset holdings of just
US$0.5 billion). Thus, to the extent that the non-financial sector did not have a natural
FX hedge (i.e., were not exporters), the US$31.5 billion of FX-risk of the banking system
had simply been transformed into credit risk. 7 Compounding this risk was the weak
capital structure of the corporate sector in Thailand (and in Asia, more generally), with an
average debt-equity ratio of 196.
These mismatches meant that Thailand’s vulnerability to a crisis was far greater than the
US$10 billion aggregate short-term liability position to the rest of the world would suggest.
Crisis Trigger
12. In the event, the proximate trigger of the crisis was the asset price deflation (stock
prices fell by 60 percent between mid-1996 and mid-1997, while inflation-adjusted property
prices fell by 50 percent between end-1991 and end-1997). This called into question the
creditworthiness of the non-financial sector and therefore the quality of banks’ assets,
including its FX cover. Against a background of an unsustainable current account deficit
(which had reached 8 percent of GDP in 1996), a significant real exchange rate appreciation,
and a weakening fiscal balance, pressures on the Thai baht increased during 1996 and the
first half of 1997. Of the US$38 billion of foreign exchange reserves at end-1996, the Bank
of Thailand used up some US$7 billion in foreign exchange intervention plus increasing its
FX forward and swap obligations from about US$5 billion to almost US$30 billion.
Information on the counterparties to these off-balance sheet swap operations is not available.
To the extent that these were Thai banks, this would have decreased the (on-balance sheet)
FX exposure of the banking system without implying a loss for the country as a whole. But if
they were nonresident entities, this would have meant that the country had only US$3 billion
6 This assumes that all medium-term liabilities to the external sector were denominated in
foreign currency.
7 Writing off the claims of the banking sector on the non-financial sector would, obviously,
worsen the balance sheet of the former, to US$41 billion.

- 8 -
of foreign exchange reserves, plus about US$3 billion of banks’ short-term foreign assets, to
cover some US$48 billion of short-term liabilities.
Argentina (2002)
13. Weaknesses in Argentina’s public sector balance sheet are well known. In particular,
by end-2001, foreign currency denominated public debt had reached 62 percent of GDP, and
gross financing need of the government had risen to 14 percent of GDP, while the central
bank’s gross foreign assets amounted to less than 5 percent of GDP (which was in any case
required for backing the central bank’s domestic monetary liabilities under the convertibility
law). Much less well-known are weaknesses in the private sector’s balance sheets and how
these contributed to the crisis.
Balance sheet vulnerabilities
14. In fact, private sector currency mismatches were severe, with foreign currency debt
larger (in relation to exports) than in the East Asian crisis—notoriously considered to be
“private sector-driven” crises. At end-2000, the Argentine corporate sector had borrowed
some US$37 billion externally as well as US$30 billion from the domestic banking sector—
for a total FX exposure of 194 percent of exports. 8 In part, this was because domestic banks
had to lend in foreign currency in order to narrow their own FX-exposure arising from
foreign currency deposits. As in Thailand, however, this meant that banks’ FX-risk was
being transformed into credit risk on households and corporations—neither of which had
significant natural hedging opportunities as Argentina’s export sector was small, and
households were using the loans for home mortgages.
15. Argentina also lost more reserves in 2001 as a result of a bank run than as a result of
the government’s inability to access external markets for its financing needs. This was
because the relatively long maturity of the government’s debt limited the pace at which
international investors could withdraw, while convertibility allowed depositors to withdraw
peso deposits and convert them into dollars. Of course, this run from peso deposits was not
unrelated to the public sector’s funding difficulties—not least because depositors recalled
how previous crises had resulted in deposit freezes (which, indeed, later happened in
the 2002 crisis as well).
16. Table 3 presents a simplified balance sheet of the banking sector. The balance sheet
provides two important insights:
• During 2001, domestic deposits and external liabilities fell by some US$24 billion,
requiring the banking system to reduce its lending to the private sector by some
8 By contrast, the corresponding exposure in terms of exports was 160 percent in Thailand
and 60 percent in Korea.

- 9 -
US$12 billion, run down liquid assets by US$5 billion, and borrow some US$9 billion
Table 3. Argentina: Principal Assets and Liabilities of the Banking System
(in billions of U.S. dollars)
End-1998 End-1999 End-2000 End-2001
Principal assets
Cash and liquid assets 8.4 8.4 8.3 3.4
Domestic currency 2.9 2.8 2.5 1.9
Foreign currency and liquid assets 5.5 5.6 5.9 1.5
Loans to and securities issued by the public sector 23.5 28.2 28.7 30.1
Domestic currency 4.8 5.5 3.7 3.4
Foreign currency 18.7 22.7 25.0 26.7
Loans to and securities issued by the private sector 70.5 68.4 65.8 54.2
Domestic currency 26.9 25.9 25.0 15.0
Foreign currency 43.7 42.5 40.9 39.1
Subtotals
Domestic currency assets 34.5 34.1 31.1 20.2
Foreign currency assets 68.0 70.9 71.9 67.4
102.5 105.0 102.9 87.6
Principal liabilities
Deposits 77.3 79.9 83.2 67.3
Domestic currency 37.3 35.8 34.7 21.7
Foreign currency 40.0 44.2 48.5 45.6
External obligations 21.4 22.8 24.1 16.3
Domestic currency 0.5 0.5 0.4 0.1
Foreign currency 20.9 22.2 23.7 16.2
Subtotals
Domestic currency liabilities 37.8 36.3 35.1 21.7
Foreign currency liabilities 60.9 66.4 72.2 61.8
98.7 102.7 107.3 83.5
Central bank support 0.3 0.2 0.1 9.2
Domestic currency 0.3 0.2 0.0 4.1
Foreign currency 1/ ... ... 0.1 5.1
Liabilities, including liabilities to central bank 99.0 103.0 107.5 92.7
Source: Table 4.2 of Rosenberg et al. (2005). Central Bank of Argentina presentation based
on Lagos (2002).
1/ Data from Lagos (2002). Central Bank of Argentina (BCRA) swap obligations disaggregated from other
obligations due to financial intermediation in BCRA data.

- 10 -
from the central bank. Moreover, as domestic currency deposits fell more rapidly than
foreign currency deposits, banks had to reduce their domestic currency-denominated
lending faster than their FX-denominated lending, even though domestic currency loans
were more likely to perform in the event of a devaluation, as opposed to FX-denominated
loans, which were likely to turn non-performing. (See Figure 2 for an illustration of
maturity mismatches, including and excluding domestic foreign currency deposits.)
Figure 2. Argentina: Maturity Mismatches: With and Without
Foreign Currency Deposits, 2001
(in billions of U.S. dollars)
Maturity Mismatch in
External Position
32.8
32.8
Maturity Mismatch
in Foreign Currency
(Including Domestic
Foreign Currency
Deposits)
-8.2
-41.0
Liquid assets
Short-term liabilities
Mismatch
-89.5
-56.7
Source: Figure 4.1 of Rosenberg et al. (2005). Country authorities and Fund staff
estimates.
• The balance sheet also shows the banking sector’s exposure to the government, with
credit to the private sector representing 28 percent of bank’s assets at the end of 2000,
and 35 percent of its FX-denominated assets. But far from being a source of strength to
the banking sector facing a deposit run, the government—facing its own gross financing
needs of some US$37 billion—was a source of weakness. The government could not
draw on the central bank’s reserves to meet its financing needs, as these were required to
back the central bank’s monetary liabilities, so the government had to turn to banks both
to roll over its maturing debts and to provide additional financing. This meant that banks

- 11 -
could not reduce their exposure to the government to meet the deposit outflow without
triggering a government funding crisis, and instead had to run down their own external
assets—the one asset that would have continued to perform in the event of default and
devaluation. Banks also had to reduce their domestic currency lending to the private
sector, though these were more likely to perform in the event of a devaluation.
Crisis trigger
17. Argentina’s experience illustrates how currency and maturity mismatches in the
public and private sector balance sheets can interact to exacerbate vulnerabilities. But what
triggered the crisis? In contrast to some other capital account crises (e.g., Uruguay) where a
specific event triggered the crisis, Argentina’s 2002 crisis was the culmination of a prolonged
period over which it became increasingly apparent that fiscal policy was not consistent with
the pegged exchange rate under the currency board arrangement regime. Traditional currency
crisis models would suggest that if the central bank expands domestic credit at a faster rate
then the growth in money demand, then the exchange rate peg will eventually collapse.
However, this was not the case in Argentina, where the central bank largely remained within
the strictures of the currency board regime. Nevertheless, Argentina’s fiscal policy was
intertemporally inconsistent with its exchange rate peg.
18. In particular, the “fiscal theory of price determination” emphasizes the intertemporal
budget constraint of the consolidated public sector (including the central bank), whereby the
nominal stock of liabilities—outstanding government debt and base money stock—deflated
by the price level must equal the present value of primary surpluses and seignorage. 9
Assuming that the public sector does not repudiate its obligations (either bonds or base
money), the intertemporal budget constraint must be satisfied. But there are two ways in
which this may happen. In a money dominant regime, the price level is determined, and it is
the stream of primary surpluses on the right-hand-side of the equation that must adjust to
maintain the government’s solvency. In a fiscal dominant regime, the stream of future
primary surpluses is given, and it is the price level that must adjust to ensure that the
government’s present value budget constraint is satisfied.
9 In mathematical terms:
∞
D ( )
t
+ M ⎧ s
t
t+ j
+ θt+
j ⎫
= E
t ⎨∑ j ⎬
(1)
Pt
⎩ j=
0 (1 + r)
⎭
where D t is the nominal stock of outstanding government debt inherited at the beginning of
period t, M t is the nominal stock of money (net of the central bank’s foreign exchange
reserves and credit to the economy) inherited at the beginning of period t, P is the price level,
s is the primary surplus and θ is central bank seignorage (in real terms), (1+r) is the
economy’s discount factor, and E{ •}
is the expectations operator.

- 12 -
19. Under a pegged exchange rate, the domestic price level is determined by the
exchange rate (for instance, by purchasing power parity or—more generally—by the
requirement that the exchange rate not become uncompetitive) and cannot, in general, adjust
to satisfy the intertemporal budget constraint. Therefore, to be viable, an exchange rate peg
requires that macroeconomic policies operate under a “money dominant” regime.
Argentina’s 2002 crisis came about as it became increasingly apparent that the country was
in a fiscal dominant regime such that the requisite fiscal surpluses were unlikely to be
generated to satisfy the public sector’s intertemporal budget constraint. 10
Turkey (2000/01)
20. In late-2000 and early 2001, Turkey suffered twin banking and balance of payments
crises when it was about ten months into an exchange-rate based disinflation program. The
disinflation program had been intended to tackle the unsustainable public debt dynamics
which had resulted in a public sector borrowing requirement of 20 percent of GDP (and a
debt ratio of 60 percent of GDP) with inflation averaging 80 percent during the 1990s.
Balance sheet vulnerabilities
21. A significant share of the public debt was in foreign currency or in short-term
domestic currency denominated Treasury bills, partly held by foreign investors. But while
weaknesses of the public sector’s balance sheet were well known, the banking system also
had a highly vulnerable balance sheet. First, because of the history of high inflation, the
average maturity of local currency deposits was short, and half of its deposits were in foreign
currency. Second, the public sector’s large borrowing requirements had crowded out the
private sector, with more than half of banks’ assets being claims on the public sector
(Figure 3).
22. The state banks’ balance sheets had massive maturity mismatches. Forced to extend
subsidized credits, they accumulated receivables from the government (“duty losses”),
requiring them to borrow heavily at short-term from households and, later in 2000, in the
overnight market to meet their liquidity needs.
23. Meanwhile, private banks were running large currency mismatches for the “carry
trade” of borrowing at low cost abroad and investing in high yield local currency government
treasury bills. The pre-announced exchange rate crawl—integral to the disinflation
strategy—provided further incentive for this arbitrage. While there were limits (15 percent of
capital) on the open FX-position that banks were allowed to run, much of the banks’ cover
was in the form of forwards with other Turkish banks or claims on domestic residents that
did not have natural hedges. Excluding such cover, the open FX position on the eve of the
crisis is estimated to have been some 300 percent of bank capital (Figure 4). The initial
10 Indeed, given Brazil’s devaluation in early 1999, the equilibrium price level in Argentina
(at a constant nominal exchange rate) had fallen, making it even more difficult to satisfy the
budget constraint (1) in footnote 9 above.

- 13 -
success of the disinflation program in lowering nominal and real interest rates also
encouraged banks to buy longer-term fixed-rate government bonds to “lock-in” the high
interest rates, but as they continued to fund themselves mainly with short-term deposits and
the overnight “repo” market, banks’ maturity mismatch worsened as well.
100
Figure 3. Turkey: Banking Sector Assets
(in percent)
Claims on the government
Claims on the private sector
80
60
40
20
0
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
Source: Figure 4.10 of Rosenberg et al. (2005). Country authorities and IMF staff
estimates.
24. Overall, therefore, the banking system’s balance sheet was highly vulnerable to an
interest rate or exchange rate shock, as banks were borrowing short-term in foreign currency
and lending in local currency to the government at increasingly long maturities.
Domestically, banks were borrowing short term and also lending at much longer term to the
government. However, given the combined public and banking sector balance sheet
mismatches, policy options were limited. The government could have decreased banks’
currency mismatch by issuing FX-denominated bonds (as it subsequently did) but at the cost
of increasing its own currency mismatch. 11 On the other hand, if banks had sought to rapidly
11 During Brazil’s 1999 currency crisis, the government increased its own currency mismatch
to help protect the banking system but had substantial foreign currency reserves that enabled
(continued…)

- 14 -
reduce their currency mismatches either by reducing FX-denominated liabilities or by
acquiring other foreign assets, this would have resulted in higher interest rates which would
not only undermine the government’s debt sustainability but also create losses for banks that
had maturity mismatches.
0
Figure 4. Turkey: Banks' Net Open Foreign Currency
Positions
(in billions of U.S. dollars)
-5
-10
-15
Excluding forwards
Including forwards and foreigncurrency-indexed
assets
-20
Jan. 2000 Sep. 2000 Dec. 2000 Mar. 2001 Dec. 2001 Dec. 2002
Source: Figure 4.12 of Rosenberg et al. (2005). Country authorities and
IMF staff estimates.
Crisis trigger
25. The crisis occurred in November 2000 amidst uncertainty about the government’s
will to tackle politically sensitive bank restructurings and against a backdrop of a widening
current account deficit (which had reached 7 percent of GDP) and a substantial real exchange
rate appreciation as inflationary dynamics—though sharply slowing—outstripped the preannounced
rate of crawl. The exodus of foreign funds led to a spike in interest rates, which
caused a drop in the value of banks’ holdings of fixed-rate government bonds as well as
increasing their funding costs. When the peg was abandoned in February 2001—following a
further exodus triggered by a political crisis—banks’ net foreign currency exposure was
revealed. While the fragility of the public sector’s balance sheet had contributed to the crisis,
it to do so. The strategy worked in that the economic impact of the subsequent devaluation
was one of the mildest among capital account crises.

- 15 -
in the aftermath its balance sheet deteriorated significantly. First, the share of domestic debt
at floating rates rose as investors demanded protection against further interest rate increases
(and banks sought to reduce the maturity mismatch between short-term deposits and longerterm
government bonds). Second, in an effort to avoid a collapse of the banking system, the
government provided a blanket guarantee for banks’ liabilities and issued bonds for their
recapitalization. These bonds increased public debt by 30 percent of GDP to almost
90 percent of GDP at end-2001.
III. IDENTIFYING BALANCE SHEET VULNERABILITIES
26. The examples above illustrate how currency and maturity mismatches in sectoral
balance sheets, and linkages between them, can contribute to the likelihood that a capital
account crisis could be—and ultimately was—triggered. At the same time, given emerging
market countries’ limited ability to borrow in their own currencies (“original sin”), there
must be FX-exposure in some sectoral balance sheet in the economy. This also means that
any “hedging” will either be incomplete or that, in effect, the country is not a net recipient of
capital from the rest of the world. Therefore, the key to reducing vulnerability is to try to
limit currency, maturity, and capital structure mismatches and ensure that risks—including to
real shocks—are ultimately contained by strong balance sheets within the economy. 12
27. Although balance sheet analysis is still in its infancy, the examples cited above
suggest some conclusions:
• The banking system often acts as a key transmission channel of balance sheet problems
from one sector into another. If a shock in the corporate sector (Asian crisis countries) or
the public sector (Russia 1998, Turkey 2001, Argentina 2002) results in it being unable to
meet its liabilities, then another sector—typically the banking sector—loses its claims. In
turn, this can cause a deposit run, sparking a banking crisis, especially if the
government’s own balance sheet is too weak to provide credible deposit insurance or
lacks international reserves to provide liquidity support in foreign exchange. By the same
token, if banks tighten their lending to prevent their portfolios from deteriorating, then
this further complicates the situation of the corporate or public sector that is facing
financing difficulties.
• If the government’s balance sheet is sufficiently strong, it can serve as a “circuit
breaker,” halting the propagation of shocks across domestic balance sheets. In a number
of recent crises (e.g., Argentina 2002), however, the government balance sheet was the
12 To use an analogy, lightning strikes might leave a house at risk of burning down and while
measures can be taken to reduce that risk (e.g., installing a lightning conductor), some risk
may be unavoidable. By purchasing insurance, however, the homeowner transfers the
associated financial risk from his own relatively weak, undiversified balanced sheet to that of
the insurance company, which is much stronger in that it holds diversified risks.

- 16 -
main source of weakness, precluding such a role. Indeed, banks typically want to hold
government securities as they may be the only liquid, domestic-currency denominated
assets. However, if—as in Argentina—the government defaults on its debt, then this can
be a source of vulnerability to the banking sector. 13
• Available foreign exchange reserves or contingent financing may be especially valuable
in reducing the economy’s balance sheet vulnerabilities as they can be used to cover
short-term financing needs of the public sector, to provide a partial lender of last resort
function in dollarized economies, or to help close the private sector’s foreign currency
mismatch—insulating the economy from the impact of a devaluation—by providing
liquidity to banks. However for contingent financing to be useful, it must be very quickly
accessible.
• Maturity and currency mismatches are sometimes hidden in indexed or floating rate
instruments. For instance, in Brazil, liabilities may be formally denominated in local
currency but linked to the exchange rate. 14 Likewise, an asset may have a long maturity
but carry a floating interest rate. Such indexation often creates the same mismatches as if
the debt were denominated in foreign currency or as if the maturity were as short as the
frequency of the interest rate adjustments.
• As was the case both in Thailand and in Argentina, balance sheet linkages can transform
one type of risk into another without necessarily reducing that risk. For example, the
banking system may try to close its FX mismatch on foreign currency deposits by lending
to domestic corporations in foreign currency. However, if the non-financial sector
recipients of those loans do not have natural hedges (e.g., have export revenues), then the
banking system’s currency risk is simply transformed into credit risk.
• Off-balance sheet items can substantially alter the overall risk exposure—reducing or
increasing balance sheet exposures according to whether an underlying position is being
hedged or the entity is taking a speculative position in the derivatives markets. However,
such transactions can also mask vulnerabilities, for instance as risk from a balance sheet
mismatch is transformed into counterparty risk. In aggregate, a sectoral balance sheet
may appear hedged through the derivative markets but may still be exposed to the risk if
13 This suggests that, when the government’s balance sheet is relatively weak, multilateral
organizations could usefully issue debt denominated in emerging market country currencies,
thus providing a domestic-currency denominated asset to the banking sector without the
corresponding default risk. Multilateral organizations would, however, assume the
corresponding currency risk.
14 Over the past couple of years, the Brazil government has gradually eliminated much of its
foreign currency-indexed debt.

- 17 -
the counterparties are connected. 15 For example, in Turkey, the banking system open FX
exposure was small when forward transactions were included, but the main
counterparties in these forward transactions were other Turkish banks.
• The ultimate buffer for private sector balance sheet mismatches (e.g., currency/FX) is
capital. A major source of vulnerability in the East Asian crises was the very high debtequity
ratios (Table 4).
Table 4. Average corporate debt-to-equity ratios in selected countries
(in percent)
Thailand
Taiwan Province
of China
United States Germany Malaysia Japan Korea
196 90 106 144 160 194 317
Source: Table 3, Annex II, of Allen et al. (2002).
• Pegged exchange rate regimes, by offering an implicit exchange rate guarantee, might
encourage greater risk taking in the form of open (mismatched) FX-positions. As noted
above, to the extent that emerging market countries’ ability to borrow in their own
currency is limited, there must be aggregate foreign currency exposure associated with
foreign liabilities (i.e., obligations to non-residents). Nevertheless, there are at least two
ways in which pegged exchange rates might exacerbate foreign currency risk:
• The implicit guarantee might encourage more “carry trade” (arbitrage between
low-cost foreign currency borrowing and higher domestic interest rates at a
given exchange rate) resulting either in greater total foreign borrowing or a
bias towards shorter maturity foreign liabilities (Thailand 1997), Turkey
2001/02).
• Again by providing an implicit exchange rate guarantee, the pegged exchange
rate might encourage more domestic “dollarization”—i.e., holding of foreign
currency-denominated assets and liabilities by residents, though neither logic
nor empirical evidence particularly supports this. 16
15 For example, a bank may be closing its spot FX exposure through a derivative transaction
with its parent conglomerate; such practices apparently occurred in Turkey prior to the 2000
crisis.
16 As pointed out in Lessons from the Crisis in Argentina (IMF Occasional Paper No. 236),
the exchange rate guarantee implicit in a pegged regime (or currency board) cannot
simultaneously explain both asset and liability dollarization. For instance, if the peg is
credible, households may want to borrow in foreign currency (since FX interest rates are
(continued…)

- 18 -
IV. TOWARDS CRISIS PREVENTION
28. The discussion above suggests where balance sheet vulnerabilities might lurk and
how they may interact with specific triggers that result in a full blown crisis. The first step in
crisis prevention is to try to avoid such vulnerabilities—in particular, to ensure that the
government is not (perhaps inadvertently) providing incentives that exacerbate balance sheet
mismatches. It is a truism that sound macroeconomic policies also lessen—but do not
eliminate—the possibility that a crisis will be triggered.
29. What can the Fund do to prevent crises? Surveillance is certainly at the heart of any
response in that regard (see Box 1). While Fund-supported programs are usually thought of
in the context of crisis resolution, recent analytical work at the IMF—Ramakrishnan and
Zalduendo (2006)—has examined a possible role in the context of crisis prevention as well.
30. What factors might determine whether a crisis is triggered? The analysis considers
the experience of 27 emerging market countries over the period 1994-04 and identifies 32
episodes of “high market pressure” (i.e., when the real exchange rate was depreciating, the
country was losing foreign exchange reserves, or sovereign bond spreads were widening). Of
these 32 episodes, 11 turned into capital account crises while the other 21 did not (Table 5).
31. The intriguing question is why those 11 cases—and not the others—turned into
crises. Part of the answer is presumably that the balance sheet vulnerabilities were more
acute in the crisis cases. However, a full comparison between the balance sheet
vulnerabilities in the 32 episodes was beyond the scope of the study. 17 Nevertheless, it is
noteworthy that the crisis countries had significantly higher external debt and short-term
debt-to-reserves ratios than the countries that managed to avoid the crisis despite the high
market pressure episode.
typically lower and there is little risk of a devaluation) but then they would not want to hold
dollar deposits. Conversely, if there are doubts about the viability of the peg, households
would want to hold dollar deposits but not borrow in foreign currency. Empirically, there
does not seem to be any association between pegged exchange rate regimes and dollarization
of the banking system.
17 Comparisons across episodes about the susceptibility of the country to a crisis are also
difficult because the balance sheet vulnerability typically interacts with a specific crisis
trigger.

- 19 -
Box 1. Surveillance at the IMF
As described by Lane (2005), greater emphasis in surveillance has been placed on
crisis prevention. Efforts to that end include consideration of both stock and flow
imbalances (the former as part of the balance sheet approach), better financial sector
surveillance, and a more systematic debt sustainability analysis (DSA). Early warning
of possible external imbalances is being attempted through regular vulnerability
exercises, established in 2001, which provide cross-country assessments of underlying
weaknesses in economic fundamentals as well as near-term crisis risks. Financial
sector surveillance and adherence to international standards in various areas have been
improved through the use of the Financial Sector Assessment Program (FSAP),
integration of financial sector issues in the Article IV consultations with member
countries, as well as Reports on Standards and Codes (ROSCs). 18 Additionally, greater
emphasis on transparency, including publication of Fund documents and subscription
to the Special Data Dissemination Standard (SDDS), has facilitated the flow of timely
information to the market, perhaps limiting adverse self-fulfilling expectations. Debt
sustainability assessments—required of all Article IV consultation reports—provide a
consistency check on baseline medium-term projections, and further identify possible
medium-term vulnerabilities.
18 As highlighted by the McDonough Commission report, further progress could still be
made in this area. The Managing Director’s Medium Term Strategy also puts a premium on
further strengthening the Fund’s financial sector surveillance capabilities.

- 20 -
Table 5. Classification of Capital Account Crises (KAC) and Control Group (CG) Episodes
Episode
Country
Beginning date of market
pressure
Identifying Market Pressures 1/
End date of market
pressure
Duration of
pressure in
months 3/
Number of
months with
pressure
KAC or CG
Episodes 2/
1 Argentina 2001 July 2002 May 11 6 KAC
2 Brazil 1998 August 1999 January 6 3 KAC
3 Bulgaria 1996 May 1996 May 1 1 KAC
4 Ecuador 2000 January 2000 January 1 1 KAC
5 Indonesia 1997 October 1998 January 4 3 KAC
6 Korea 1997 October 1997 December 3 3 KAC
7 Malaysia 1997 July 1998 January 7 5 KAC
8 Russia 1998 August 1998 September 2 2 KAC
9 Thailand 1997 July 1997 August 2 2 KAC
10 Turkey 2000 November 2001 March 5 3 KAC
11 Uruguay 2002 July 2002 July 1 1 KAC
1 Argentina 1998 August 1998 August 1 1 CG
2 Brazil 2002 July 2002 July 1 1 CG
3 Bulgaria 1998 August 1998 August 1 1 CG
4 Chile 1999 June 1999 June 1 1 CG
5 Chile 2002 June 2002 June 1 1 CG
6 Colombia 1998 April 1998 September 6 3 CG
7 Colombia 2002 July 2002 August 2 2 CG
8 Hungary 2003 June 2003 June 1 1 CG
9 Indonesia 2004 January 2004 January 1 1 CG
10 Mexico 1994 December 1995 March 4 3 CG
11 Mexico 1998 August 1998 August 1 1 CG
12 Peru 1998 August 1998 December 5 2 CG
13 Philippines 1997 August 1997 August 1 1 CG
14 Poland 1998 August 1998 August 1 1 CG
15 South Africa 1996 April 1996 April 1 1 CG
16 South Africa 1998 July 1998 July 1 1 CG
17 South Africa 2001 December 2001 December 1 1 CG
18 Turkey 1998 August 1998 August 1 1 CG
19 Venezuela 1994 June 1994 June 1 1 CG
20 Venezuela 1998 August 1998 August 1 1 CG
21 Venezuela 2003 January 2003 January 1 1 CG
Source: Table 1 of Ramakrishnan and Zalduendo (2006).
1/ Market pressures identified by classifying monthly data into five clusters based on an index of market pressures
that includes changes in REER, FX reserves, and spreads. The listed countries are in the cluster with the highest market pressures.
2/ Private capital flows (net of FDI) is used for distinguishing between KAC and CG episodes. A KAC event requires two
quarters of either medium outflows or high outflows (as defined by cluster analysis) in the four quarters that follow the
build-up of market pressures. All other episodes are in the control group (CG).
3/ Numbers of months from the beginning to the end of each market pressure episode.
32. Their econometric analysis (discussed in IMF 2006) shows that:
• Less flexible exchange rate regimes are associated with a higher likelihood that a market
pressure event turns into a crisis and an overvalued exchange rate (relative to trend) is
significantly associated with a higher likelihood of a crisis.
• Lower external debt (as a percent of GDP) is significantly associated with a lower
likelihood of a crisis.

- 21 -
• A higher stock of foreign exchange reserves (as a percent of reserves) is significantly
associated with a lower likelihood of a crisis.
• Stronger policies—tighter monetary policy or greater fiscal adjustment (particularly in
the context of a Fund-supported program)—are significantly associated with a lower
likelihood of a crisis.
• An on-track IMF-supported program is associated with a lower likelihood of a crisis, but
the effect is not statistically significant.
• Availability of Fund resources is a significant factor in crisis prevention: the larger are
the available Fund resources (as a share of short-term debt), the lower is the likelihood of
a crisis.
33. These results suggest that there is an important liquidity effect of Fund support on
crisis prevention since it is the availability of Fund resources (disbursements or their
availability for drawing under an on-track precautionary program) that matters, rather than
just an on-track program or possible future drawings under the arrangement.
34. The benefits of Fund support go beyond the liquidity effects, however, since the
available Fund financing variable is significant even controlling for the country’s available
foreign exchange reserves. Part of the effect must thus arise from a combination of stronger
policies (i.e., beyond the fiscal balance and real interest rates included in the regressions)
bolstered by conditionality and in the “seal of approval” implicit in Fund disbursements.
Moreover, since the program dummy is not statistically significant, but the Fund financing
variable is strongly significant, the strength and the credibility of the Fund’s signal appears to
depend at least to some degree on the extent to which the Fund is willing to put to its own
resources on the line.
35. Finally, it bears emphasizing that the interaction of limited currency and maturity
mismatches (low external debt-to-GDP and low short-term debt-to-reserves ratios), strong
policies, and IMF financing is critical for crisis prevention. If there are large balance sheet
mismatches and weak policies, not only is there a high probability of a crisis, the marginal
impact of IMF financing on lowering the probability of crisis is also small—thus the country
would be highly vulnerable to a crisis (Figure 5).

- 22 -
Figure 5. Marginal Impact of Fund Financing, Given Country Fundamentals 1/
Probability
of crisis
1.00
0.80
0.60
0.40
Average Fund financing among
KAC episodes
Maximum level of Fund financing
among KAC episodes
P(crisis; best
covariates)
P(crisis; median
covariates)
0.20
0.00
0.00 0.05 0.10 0.15
Fund financing
(as share STD)
P(crisis; worst
covariates)
1/ B d i 4 i T bl 2 F d fi i i d fi d h l i di b 12
Source: Figure 7 of Ramakrishnan and Zalduendo (2006).
1/ Based on Regression 4 in Table 2 of Ramakrishnan and Zalduendo (2006). Fund financing
is defined as cumulative disbursements over 12 months as a share of short-term debt. The
figure reflects the probability of crisis for different countries based on covariate contributions
at time t-1. Vertical lines are also measured at t-1 and represent, respectively, the average and
maximum level of Fund financing among crisis episodes.
V. CONCLUSIONS
36. For most emerging market countries, current market conditions are exceptionally
benign with spreads almost an order of magnitude lower than just a few years ago. Yet recent
events have also shown that these countries remain susceptible to shifts in market sentiment.
Therefore, the currently benign conditions should not breed complacency but instead provide
some breathing space for countries to address existing vulnerabilities. 19
37. Most capital account crises appear to have been caused by foreign currency and
maturity mismatches on private or public sector balance sheets coupled with a specific
trigger—domestic or external. Based on the experience of these countries, this paper has
sought to identify where and how such balance sheet vulnerabilities might arise.
38. Turning to factors that determine whether a crisis will occur, empirical analysis
suggests that minimizing balance sheet mismatches (a low external debt ratio, a low shortterm
debt-to-reserves ratio), strong macroeconomic policies, and avoiding overvaluation of
the exchange rate, contribute to reducing the likelihood of a crisis. Given that holding foreign
exchange reserves is costly, a particularly interesting result is that IMF resources disbursed
(or available under a precautionary program) have an even larger impact on crisis prevention
than the country’s own reserves. This probably reflects a combination of stronger policies
19 For example, over the past couple of years, Brazil has been reducing the foreign currency
exposure of its public sector balance sheet.

- 23 -
under an IMF-supported program, the greater credibility of the authorities’ policies, and the
stronger signal to markets of the IMF putting its own resources on the line.

- 24 -
References
Allen, M., et al., 2000, “A Balance Sheet Approach to Financial Crisis,” IMF Working Paper
No. 02/210 (Washington: International Monetary Fund).
Daseking, C., et al., 2005, Lessons from the Crisis in Argentina, IMF Occasional Paper
No. 236 (Washington: International Monetary Fund).
Flood, R., and P. Garber, 1984, “Collapsing Exchange-Rate Regimes: Some Linear
Examples,” Journal of International Economics, Vol. 17 (August), pp. 1-13.
International Monetary Fund, 2006, Fund Supported Programs and Crisis Prevention,
available at http://www.imf.org/external/np/pp/eng/2006/032306.pdf.
Krugman, P., 1979, “A Model of Balance-of-Payments Crises,” Journal of Money, Credit
and Banking, Vol. 11, No. 3, pp. 311-25.
Lagos, M., 2002, “The Argentine Banking Crisis 2001-2002,” Report prepared for the
Argentine Banking Association.
Lane, P., 2005, “Tensions in the Role of the IMF and Directions for Reform,” World
Economics, Vol. 6, No. 2, pp. 1-20.
Lipschitz, L. et al., 2005, “Real Convergence, Capital Flows, and Monetary Policy: Notes on
the European Transition Countries,” pp. 61-69 in Schadler, S., ed., Euro Adoption in
Central and Eastern Europe: Challenges and Opportunities (Washington:
International Monetary Fund).
Obstfeld, M., 1994, “The Logic of Currency Crises,” Cahiers Economique et Monetaires
No.43, pp. 189-213 (Paris: Banque de France).
Ramakrishnan, U., and J. Zalduendo, 2006, “The Role of Fund Support in Crisis Prevention,”
IMF Working Paper No. 07/75 (Washington: International Monetary Fund).
Rosenberg, C., et al., 2005, Debt-Related Vulnerabilities and Financial Crises, IMF
Occasional Paper No. 240 (Washington: International Monetary Fund).
Roubini, N., and B. Setser, 2005, Bailouts or Bail-ins? Responding to Financial Crises in
Emerging Economies (Washington: Institute for International Economics).